Interview: ParTec Turns Clusters into True Supercomputers

Print Friendly, PDF & Email

Hugo Falter is widely known in the industry for his work with the European Open File System consortium and the Lustre Community, but his company, ParTec may not be so familiar to many of us in the States. As part of our continuing series on European HPC vendors, I caught up with Hugo to discuss ParTec, EOFS, and what’s in store for ISC’12.

insideHPC: Who is ParTec and who do you help?

Hugo Falter: ParTec Cluster Competence Center GmbH is a spin-off from the “Karlsruhe Institute of Technology” (KIT) in 1999. The main purpose of the company is to enable compute – cluster to run as a Supercomputer.

Based on the ParaStation Software Suite, ParTec is contributing with our own MPI, Communication and Process Management (open source) as well as the graphical interface Grid Monitor and the analysis tool Health Checker to the highest performance, scalability and productivity of clusters.

ParTec also offers support and services together with the ParaStation Software Suite.

insideHPC: Can you tell us more about the Parastation open source project and how it fits into ParTec’s HPC effort?

Hugo Falter: ParTec is co-developing the software “ParaStation” within the “ParaStation – consortium” as an open-source initiative together with the Forschungszentrum Jülich (Jülich Research Center) and the Karlsruhe Institute of Technology (KIT) to contribute out of Europe to the international efforts to enable compute clusters to run beyond petascale and reach exascale by the end of this decade.

insideHPC: Can you describe ParTEc’s biggest HPC installation?

Hugo Falter: ParTec´s biggest HPC – installation is the cluster “JuRoPA 2” in the “Jülich Supercomputing Centre” in Germany. The system was #10 in the TOP 500 list in June 2009 and the most productive Supercomputer based on commodity components (cluster).

insideHPC: Do you derive most of your business from extreme-scale installations like JuRoPA?

Hugo Falter: ParTec Cluster Competence Center´s main focus is supercomputing with clusters.

It is crucial for systems with commodity components to have the right software to scale and deliver highest performance and productivity. But this is not enough: the system administrators who work with the supercomputer need a team of experienced technicians to  keep the system in shape. ParTec´s support team provides such necessary experience.

This helps also customers with small and midsize systems.

insideHPC: Data-intensive computing seems to be experiencing unprecedented growth. How does ParTec help customers deal with these kinds of challenges?

Hugo Falter: ParTec is the co-founder of EOSF (European Open File-Systems Cooparative). We are aware of the fact, that without a performant and highly scalable I/O – system there would be no way to to handle the ever growiing amount of data.

This is why ParTec is involved in initiatives to further develop Lustre and give room for research for highest scalable parallel file-systems.

insideHPC: Why has ParTec chosen to be a vendor-neutral integrator? What advantages does that offer your customers?

Hugo Falter: The need to support the customer is independent of the hardware vendor. For this reason ParTec is vendor neutral.

insideHPC: ISC’12 is approaching fast. What does ParTec plan to showcase there at the exhibition?

Hugo Falter: ParTec will show that the “ParaStation Software Suite” and the experience of our support team is needed to enable a cluster to run as a supercomputer.